International Journal of Heat and Mass Transfer 48 (2005) 11071116

www.elsevier.com/locate/ijhmt

Heat transfer properties of nanoparticle-in-uid

dispersions (nanouids) in laminar ow
Ying Yang a, Z. George Zhang b,*, Eric A. Grulke
William B. Anderson b, Gefei Wu b
b

a,*

a
Department of Chemical and Materials Engineering, University of Kentucky, Lexington, KY 40506, USA
New Product Development Laboratory, The Valvoline Company, P.O. Box 14000, Lexington, KY 40509, USA

Received 12 February 2004; received in revised form 15 September 2004

Available online 10 December 2004

Abstract
The convective heat transfer coecients of several nanoparticle-in-liquid dispersions (nanouids) have been measured under laminar ow in a horizontal tube heat exchanger. The nanoparticles used in this research were graphitic
in nature, with aspect ratios signicantly dierent from one (l/d
0.02). The graphite nanoparticles increased the static
thermal conductivities of the uid signicantly at low weight fraction loadings. However, the experimental heat transfer
coecients showed lower increases than predicted by either the conventional heat transfer correlations for homogeneous uids, or the correlations developed from the particle suspensions with aspect ratios close to one. New correlations on heat transfer need to be developed for nanouid systems.
2004 Elsevier Ltd. All rights reserved.
Keywords: Nanoparticles; Dispersions; Thermal conductivity; Heat transfer coecient; Laminar ow

1. Introduction
Heat transfer uids provide an environment for adding or removing energy to systems, and their ecacies
depend on their physical properties, such as thermal
conductivity, viscosity, density, and heat capacity. Low
thermal conductivity is often the primary limitation for
heat transfer uids. Recently, there has been interest in
using nanoparticles as additives to modify heat transfer

*
Corresponding authors. Tel.: +1 8593573510; fax: +1
8593573530 (Z.G. Zhang); tel.: +1 8592576097; fax: +1
8593234922 (E.A. Grulke).
E-mail addresses: zzhang@ashland.com (Z.G. Zhang),
egrulke@engr.uky.edu (E.A. Grulke).

uids to improve their performance [112]. Dispersion

or suspension of nanoparticles of high thermal conductivities in heat transfer uids (the so-called nanouids)
is one of the methods for improving the thermal conductivity of the mixtures [15], and thus increasing their
heat transfer coecient in various applications. Some
examples of nanouids with improved thermal conductivity include metal nanoparticle suspensions as working
uids in microchannel heat exchangers [1], copper oxide
particles suspended in water [3], and silicon carbide
nanoparticles in water or ethylene glycol [9,12]. Compared with millimeter- or micrometer-sized particle suspensions, nanouids possess better long-term stability
and rheological properties, and can have dramatically
higher thermal conductivities. Current eorts have
mainly been focused on low aspect ratio nanoparticle

0017-9310/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2004.09.038

1108

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

Nomenclature
A
CP
d
D
g
h
k
l
L
m_
n
q
T
U
W

area
heat capacity
diameter of particle
diameter of test tube
acceleration due to gravity
heat transfer coecient
thermal conductivity
thickness of particle
length of test tube
mass ow rate
shape factor
heat transfer rate
temperature
overall heat transfer coecient
weight ow of uid

/
Gr
Gz
Nu
Pe
Pr
Re

Subscripts
b
bulk
e
equivalent
e
eective
f
uid
hf
heating uid
i
inside
in
inlet
o
outside
out
outlet
p
particle
w
wall

Greek symbols
b
coecient of expansion
l
viscosity
m
kinematic viscosity
q
density

dispersions (disks), since controlling the thickening eect

of the high aspect ratio particles (rods) is still a very
challenging task [10,13].
A model has long been established by Hamilton and
Crosser [14] to estimate the eective thermal conductivity, ke, of macroscopic solidliquid mixtures, which is
given in the following equation:
k eff k p n  1k f  n  1/k f  k p

k p n  1k f /k f  k p
kf

volume fraction of nanoparticles

Grasho number
Graetz number
Nusselt number
Peclet number
Prandtl number
Reynolds number

in which kf and kp are the thermal conductivities of the

uid and particles, respectively. / is the volume fraction
of the particles, and n is the empirical shape factor.
Large aspect ratio particles, such as carbon nanotubes
[10], have high values of n in Eq. (1), and therefore more
potential for thermal conductivity enhancement. However, the addition of large aspect ratio particles into a
liquid may result in huge increase in viscosity as compared to the continuous phase [13].
Recent work by Choi and colleagues [4,5,8,10] indicates that the thermal conductivity increase caused by
the addition of nanometer-sized particles is much higher
than predicted by the above equation. For example, the
addition of 0.3 vol% of copper nanoparticles increases
the thermal conductivity of ethylene glycol by 40% [8],
and 1 vol% carbon nanotubes lead to a 150% increase
of the thermal conductivity of a synthetic poly(a-olen)
oil [10].
Keblinski et al. [15] explored the mechanisms of heat
transfer in nanouids, and proposed four possible rea-

sons for the contribution of the nanometer-sized particles to the increase of the thermal conductivity of the
system: Brownian motion of the particles, molecularlevel layering of the liquid at the liquid/solid interface,
the nature of the heat transport in the nanoparticles,
and the eects of nanoparticle clustering. The authors
of the current communication believe there may be a
synergistic eect of several above-mentioned mechanisms, among which the percolation eect may be
favored for particles with high or low aspect ratios.
Nanouids are multicomponent systems, and the
morphology and orientation of the dispersed solids is
complex. That is probably the reason that very few correlations have been developed for their convection heat
transfer coecients [16,17]. Considering the small size
and the low volume fraction of the particles in most
nanouids, it might be reasonable to treat nanouids as
pure liquids in certain cases. Under these circumstances,
the conventional correlations for homogeneous liquids
might be applied to these systems. Three such correlations are the SeiderTate equation for convective heat
transfer of laminar ow in tubes [18], the Oliver correlation [19], and the Eubank and Proctor correlation [20].
The SeiderTate equation is [18]
Nu 1:86Re

1=3

Pr

1=3

 1=3  0:14
D
lb
lw
L

in which Nu is the Nusselt number, Re is the Reynolds

number, and Pr is the Prandtl number. llwb 0:14 repre-

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

sents the radial variation of uid properties and natural

convection eect.
The Oliver correlation [19] for horizontal ow in
tubes is
 0:14
l
Nu w
1:75Gz 5:6  104 Gr  Pr  L=D0:70 1=3
lb
3
WC p
,
kL

in which Gz is the Graetz number, Gz

Gr is the
bf DTD3 q2f g
Grasho number, Gr
, W is weight ow of
l2f
uid, and bf, qf, and lf are the coecients of expansion,
the density, and the viscosity of uid measured at lm
temperature, respectively.
The Eubank and Proctor correlation [20] for laminar
ow in horizontal tubes is
 0:14
l
Nu w
1:75Gz 12:6Gr  Pr  D=L0:40 1=3
4
lb
In both Eqs. (3) and (4) the Grasho number, Prandtl number, and the ratio of L/D are intended to account for the eect of the natural convection.
Only a few papers have discussed the heat transfer
coecients of nanouids in convective ows
[1,6,7,16,17]. Choi [1] pointed out that heat transfer
coecients should increase with ow rates or with
increasing thermal conductivities of the uid provided
that other properties of the nanouid system, such as
heat capacity, density, and viscosity, are kept the same
as the base uid. If a nanouid can have high thermal
conductivity at low volume fractions, a high heat transfer coecient might be obtainable without the pumping
power being increased signicantly. In heat exchanger
design, the length of the exchange surface aects the local heat transfer coecient via entrance eects and other
factors. In some cases, using nanouids can help achieve
higher exchanger eciency without increasing exchanger
size. Recently Li and Xuan [16,17] studied convective
heat transfer and ow characteristics of Cu-water nanouids. The heat transfer coecients of their nanoparticle
suspensions are much higher than that of the base uid
and, when the volume fraction of nanoparticles is low,
the friction factor of the uids is not changed. They gave
the new correlation for the convection heat transfer
coecient of nanouids in a horizontal tube as [17]:
Nu 0:43281:0 11:285/0:754 Pe0:218 Re0:333 Pr0:4

in which Pe is the particle Peclet number.

The purpose of this article is to relate increases of
thermal conductivity with the increases of the heat transfer coecient for nanouids containing low aspect ratio
nanoparticles. The data are compared to Eqs. (2)(5) to
determine whether the correlations developed for
homogenous uids could be applied to the nanoparticle
dispersions with aspect ratios dierent from one.

1109

2. Experimental
Two series of nanouids are investigated in this study
with dierent base uids. The base uid for Series 1
(abbreviated as BF#1 hereafter) is a commercial automatic transmission uid (ATF), while that for Series 2
(abbreviated as BF#2 hereafter) is a mixture of two
commonly used synthetic baseoils with commercial
additive packages. The experimental nanouids listed
in Table 1 are stable dispersions of base uids with
nanoparticles (they show no signs of settling after 30
days of rest).
All uids were tested at two dierent temperatures.
Comparing uid EF#1-1 and EF#1-2 illustrates the
eects of nanoparticles loading. Comparing uid
EF#1-1 with EF#2-2 illustrates the eects of base uid
dierences. Comparing uid EF#2-1 with EF#2-2 illustrates the eects of dierent nanoparticles sources. The
average diameter of the graphite nanoparticle is about
12 lm, with the thickness around 2040 nm (l/d = 0.02).
The following physical properties of the nanouids
were measured: thermal conductivity (k), density (q),
heat capacity (CP) and kinematic viscosity (m). Thermal
conductivity measurements were done by using an inhouse transient hot-wire device based on the design by
Nagasaka and Nagashima [21]. Density was measured
using a pycnometer. Heat capacity was measured using
a dierential scanning calorimetry (TA Instruments,
Model 2920 Modulated). Kinematic viscosity was measured using a reverse tube capillary viscometer.
A small volume ( 100 ml) ow loop heat transfer system has been constructed in-house. A schematic is presented in Fig. 1, with the geometry of the heat transfer
sections illustrated. Because L/D
1, it is reasonable
to neglect the entrance eects in the test system. The
whole system is heavily insulated to reduce heat loss.
Pipes are wrapped with insulation material, and plastic
ttings are attached at both ends of the test area to thermally isolate the connection. Static mixers are put at the
ends of the test section to mix the uids and improve
accuracy of the bulk uid temperature measurement.
A positive displacement pump is used to control the
uid ow rate, and its displacement settings are calibrated to determine the volumetric ows (62507 cm3/
min) that give average linear velocities between 6.3 and
Table 1
Experimental uids

EF#1-1
EF#1-2
EF#2-1
EF#2-2

Base uid

Nanoparticle

Particle loading

BF#1
BF#1
BF#2
BF#2

Graphite
Graphite
Graphitea
Graphite

2 wt%
2.5 wt%
2 wt%
2 wt%

a
From a dierent source than that used in the other
experimental uids.

1110

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116
Heating
jacket

3
4

Thermostat

Pump

Bubble collector
Cooling jacket
Thermostat

Test tube
Heating jacket

Length
(cm)
45.7
45.7

Inside diameter
(cm)
0.457
1.09

Outside diameter
(cm)
Roughness
0.635
smooth
/
smooth

Fig. 1. Heat transfer testing apparatus. (/) thermocouple.

52 cm/s. The uid rst passes through the thermostatic

counter-current ow heating jacket in which water is
used as the heating uid, with its temperature set at
either 50 C or 70 C. Four thermocouples are placed
as shown in Fig. 1, two of which, 1 and 4, measure the
uid temperature at the inlet and outlet of the test area,
and the other two, 2 and 3, are placed at the outlet and
inlet of the heating jacket to measure water temperature.
A thermostatic cooling jacket system, which is adjusted
to 5 C, is applied to the system for it to reach a steady
state. The bubble collector collects and separates air
bubbles from the uid.
The experimental procedures are as follows: (1) the
system is warmed up by starting the pump and heater.
It takes 510 min to increase the temperature to the
operating range and remove interior bubbles; (2) The
ow rate is set, then open the cooling system for temperature control; (3) The system usually reaches a thermal
steady state within 30 min (the variability of temperature
measurements is 0.1 C): (4) Each measurement was repeated at least once. Average heat transfer coecients
were calculated. The deviations of the data are less than
2% in most cases and less than 5% in some high-ow rate
cases.

3. Results and discussion

3.1. Physical properties of the test uids
Physical properties of the base uids and nanouids
are illustrated in Table 2. Thermal conductivities of
the base uids are typical for hydrocarbon uids, in
the range from 0.12 to 0.18 W/m K. The increase of the
thermal conductivity of the nanouid versus the base
uid ranges from 3% to 56%. If we apply Eq. (1) to
the nanouid with highest particle loading, EF#1-2,
and assume that the thermal conductivity of the particle
is 100 W/m K, and the empirical shape factor can be calculated from the size and shape of the particles, we
would get a increase of at most 12.5%, much lower than
the measured value, 56%. As being discussed earlier, Eq.
(1) is deduced from suspensions containing millimeter to
micrometer-sized particles, not nanometer-sized ones.
Density and heat capacity do not show much change
as compared to those of the corresponding base uid,
which is reasonable due to the low volume fraction of
the particles and moderate temperature change of the
system. Kinematic viscosity is measured at four temperatures, 35, 43, 50 and 70 C, in which the latter two are

Table 2
Physical properties of the test uids

BF#1
EF#1-1
EF#1-2
BF#2
EF#2-1
EF#2-2

Thermal conductivity

Density (g/cm3)

k (W/m K)

k/k0

35 C

43 C

0.134
0.173
0.209
0.148
0.153
0.182

1
1.29
1.56
1
1.03
1.23

0.840
0.847
0.847
0.820
0.823
0.825

0.835
0.842
0.843
0.815
0.819
0.820

Heat capacity
(J/g K)

Kinematic viscosity (cSt)

35 C

43 C

50 C

70 C

2.10
2.10
2.10
2.20
2.20
2.20

44.8
41.4
52.4
28.5
29.3
31.1

33.9
30.6
38.4
21.9
22.7
24.1

26.8
23.7
29.6
17.5
18.4
19.5

14.5
12.2
14.9
9.78
10.5
11.2

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

The inlet and outlet temperature measurements are

based on the average uid temperature taken after the
static mixers. The heat transfer rate from the heating
uid is
7

Subscript hf is used to represent the heating uid.

The dierences between energy lost by the heating uid
and energy gained by the process uid were around 25%,
which is typical for a small-scale testing equipment. The
mass ow rate of the heating uid is much higher than
that of the test uid so the value of the temperature difference, (Tout  Tin)hf, is very small (less than 1 C) and
contributes to the energy balance inaccuracies. Incomplete isolation could also contribute to energy losses in
the system. The heat obtained by the process uid was
used as the energy transferred as it was the more accurate value.
We can calculate the heat transfer coecient of the
test uid, hi, through the following equation:
1
1
Do Do 1

ln

Di ho
U hi Ai =Ao 2k

3.1.2. Nanoparticle loading eect

The eects of nanoparticle loading can be seen by
comparing EF#1-1 with EF#1-2 (Fig. 2a and b).
EF#1-1 with 2 wt% graphite nanoparticles has a heat

600

where Ao = pDoL, and DT is the log-mean temperature

dierence.
The outside heat transfer coecient ho is computed
from the Monrad and Pelton
s equation [22] for turbulent ow in annuli, Nu = 0.020Re0.8Pr1/3(D2/D1)0.53, in
which D1 and D2 are the inner and outer diameter of
the annulus respectively. The heating uid is water in
turbulent ow, so the heat transfer coecients
(3780 W/m2 K for 50 C and 4784 W/m2 K for 70 C) are
much higher than that of the process uid. The equivalent diameter is used for heat transfer calculations,

400
300
200
100

where U is the overall heat transfer coecient, k is the

thermal conductivity of the tube wall, and hi and ho
are the individual heat transfer coecients of the inside
and outside uids, respectively. U is given by
q
U
9
Ao DT

(a)

500

20

40

60

80

60

80

Re
600

(b)

500
400

qhf m_ hf C Phf T out  T in hf

3.1.1. Reynolds number eect

The experiments were conducted over a range of
Reynolds numbers, 5 < Re < 110. The base uids and
the nanouids all show increasing heat transfer coecients as the average ow velocity, and Reynolds numbers increase (Figs. 2 and 3).

h/(W/m .K)

_ P T out  T in
q mC

OD
De IDaOD
, where IDa is the inside diameter of the
annulus and OD is the outside diameter of the tube.
The thermal conductivity of base uid #1 (BF#1) is
0.134 W/m K, and its viscosity decreases by a factor of
3 over the temperature range 3570 C. Figs. 2 and 3
show the heat transfer data as a function of Reynolds
number for all test uids at dierent experimental conditions. These gures illustrate the eects of Reynolds
number, temperature, nanoparticle loading and source,
and base uid, on the heat transfer properties of nanouids. A power law correlation, h = a Reb, has been used
to represent the base uid data as an aid to the eye, with
R2 values as 0.97 for BF#1 and 0.95 for BF#2,
respectively.

h/(W/m .K)

the temperature of the heating uid (water), and the rst

two are the calculated average temperature of the test
uid at the center of the test section at the corresponding
heating uid temperature. The dependence of kinematic
viscosity on temperature is fairly signicant for all uids.
Both base uids were formulated with viscosity modiers to get their viscosity to the appropriate range required by the application. Nanoparticles themselves
modify uid viscosity, so viscosity modiers were adjusted to make sure they have similar viscosity with base
uids.
The heat transfer rate into the process uid is

1111

300
200
100
0

20

40

Re
Fig. 2. Plot of heat transfer coecient versus Reynolds number
for Series 1 uids (a) 50 C, (b) 70 C. () base uid 1; (j)
EF#1-1; (m) EF#1-2; () tting for BF#1 according to power
law correlation.

1112

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

h/(W/m2.K)

600

contact between nanoparticles, and (2) depletion of particles in the near-wall uid phase [23], leading to an
intrinsically lower thermal conductivity layer at the wall.
Understanding and isolating which mechanism or mechanisms might be responsible for the experimental results
will require signicant future work, including computational uid dynamic modeling of the ows in nanoparticle dispersions.

(a)

500
400
300
200
100
0

600

30

60
Re

90

60

90

120

(b)

h/(W/m2.K)

500
400
300
200
100
0
0

30

120

Re
Fig. 3. Plot of heat transfer coecient versus Reynolds number
for Series 2 uids (a) 50 C, (b) 70 C. () Base uid 2; (h)
EF#2-1; (n) EF#2-2; (- - - -) tting for BF#2 according to
power law correlation.

transfer coecient similar to that of the base uid

(BF#1) at both temperatures. There is a modest increase
in the heat transfer coecient, but it is not much dierent from the experimental error. At 50 C, the nanouid
(EF#1-2) with the higher loading (2.5 wt%) has a typical
increase in heat transfer coecient of 22% over the base
uid, while its thermal conductivity is about 50% higher
than that of the base uid. At 70 C, the heat transfer
coecient increase averages 15%.
The thermal conductivity ratios of the nanouids to
the base uid are about 1.30 and 1.50, suggesting that
under quiescent conditions, both nanouids are above
the percolation limit at which there is excellent particle-to-particle contact. However, the experimental results on heat transfer properties indicate that over the
Reynolds number range, 5 < Re < 80, the nanouid with
2 wt% nanoparticles loading is not above the percolation
limit, while the nanouid with 2.5 wt% loading is.
3.1.3. Temperature eect
Fig. 2a shows that, at 50 C, EF#1-2 has a typical increase in heat transfer coecient of 22% over the base
uid. However, at 70 C, the heat transfer coecient increase averages only 15%. Several mechanisms may lead
to smaller improvements in heat transfer coecient
between the nanouids and the base uids at higher
temperature. These include: (1) rapid alignment of
\nanoparticles in lower viscosity uids, leading to less

3.1.4. Nanoparticle source eect

Fig. 3a and b compare data for two dierent graphite
nanoparticles at the same loading, 2 wt%, in base uid 2.
At 50 C, the heat transfer coecient of the base uid is
modeled well by the power law expression (R2 = 0.98).
The data for dispersion EF#2-1, which contains graphite particles from a dierent source than the rest of the
nanouids, suggests that this nanouid is not above
the percolation limit for this range of Reynolds numbers
and temperatures. On the other hand, the uid with
graphite material used in the rest of the study, EF#22, shows increases in its heat transfer coecient relative
to that of the base uid at higher Reynolds numbers
(Re > 40). There may be similar eects such as nanoparticle alignment and nanoparticle separation near the wall
that are sensitive to nanoparticle structure and interaction. The dierence could be caused by particle shape,
morphology, or surface treatment. In any case, this comparison demonstrates that one type of nanoparticle is
more ecient than the other in increasing the heat transfer coecient of the nanouid.
3.1.5. Base uid eect
The choice of the base uid also aects the heat
transfer coecients of nanouids. In EF#1-1 and
EF#2-2, the same graphite nanoparticles were dispersed
in the two base uids at 2 wt%. EF#1-1 has similar heat
transfer coecients to BF#1 at both temperatures.
EF#2-2 has higher heat transfer coecients than
BF#2, particularly at higher Reynolds numbers. These
data demonstrate that Reynolds number, temperature,
nanoparticle loading, nanoparticle source, and the
choice of the base uid can all aect the heat transfer
coecients of nanouids.
3.2. Correlations for convective heat transfer
The correlations for the convective heat transfer of
the single-phase uid may be applied to predict h of a
nanouid system, if the volume fraction of particles is
very low. Eq. (2) can be re-written into a form more convenient for identifying the impact of Reynolds number
on the heat transfer coecient:
 1=3  0:14
L
lb
X Nu  Pr1=3
1:86Re1=3
10
D
lw

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

Fig. 4 shows the experimental results plotted as X

versus Re.
The 1/3-dependence as predicted by Eq. (10) is clearly
shown in Fig. 4. However, at low Reynolds number,
Re < 30, the data are more scattered. The larger error
at low Reynolds number is likely due to the eect of natural convection. As a rst approximation, Eq. (10) is
adequate, but it obviously misses important information
about the thermal properties of the nanoparticle suspension, particularly at low Reynolds numbers.
The data sets can be correlated using the following
equation:
X aReb

11

The results listed in Table 3 show that the Reynolds

number exponent, b, of the nanouid is fairly close to
that of the base uid, with the value at 70 C lower than
that at 50 C for the same uid. This eect seems to be
more apparent for the Series 2 uids (i.e., BF#2,
EF#2-1, and EF#2-2), and could be due to the fact that
the viscosity of Series 1 uids is higher. For uids with
10

1113

higher viscosity, natural convection (for example, the

term including the Grasho number in Eq. (3)) contributes less to the heat transfer eect, which results in the
smaller temperature dependence for the Series 1 uids.
The value of the pre-exponential coecient, a, of the
suspension is always lower than that of the corresponding base uid. Fig. 5 shows that most of the experimental data for the nanouids are below the theoretical line
(Eq. (10)) while the base uid values are generally the
same as the theoretical line.
Eq. (10) provides a basis for comparing the heat
transfer coecients calculated from the theory with
those calculated from the experimental data. Since density and heat capacity of the nanouid are very close to
those of the base uid (Table 1), the heat transfer coefcient ratio should be sensitive to two important terms,
k2/3 and llwb 0:14 . The comparison is listed in Table 4.
Table 4 indicates that the increase of the heat transfer
coecient of nanouid is much less than that predicted
from a conventional correlation. It seems that some

10

Base Fluids

y = 1.86 x 0.3333

1
1

10

100

1000

(a)

Re

y = 1.86 x

0.3333

1
1

Fig. 4. Plot of X versus Reynolds number for all the test uids.
() test data; () tting according to Eq. (10).

10

100

1000

Re
10

Table 3
Coecients of Eq. (11) for each test uid
Heat uid
temperature
(C)

R2 of the
tting

50
50
50
70
70
70
50
50
50
70
70
70

1.90 0.13
1.64 0.08
1.66 0.10
2.28 0.19
2.00 0.19
1.88 0.13
2.09 0.14
2.17 0.12
1.81 0.13
3.11 0.30
2.92 0.30
2.66 0.24

0.33 0.02
0.33 0.01
0.34 0.02
0.28 0.02
0.28 0.03
0.30 0.02
0.31 0.02
0.29 0.01
0.33 0.02
0.22 0.02
0.23 0.03
0.24 0.02

0.980
0.991
0.984
0.964
0.952
0.976
0.982
0.985
0.983
0.931
0.928
0.948

BF#1
EF#1-1
EF#1-2
BF#1
EF#1-1
EF#1-2
BF#2
EF#2-1
EF#2-2
BF#2
EF#2-1
EF#2-2

Nanofluids

y = 1.86 x

(b)

0.3333

1
1

10

100

1000

Re
Fig. 5. Plot of X versus Reynolds number for base uids and
nanouids. (h) test data on base uids; () test data on
nanouids; () tting according to Eq. (10).

1114

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

Table 4
Heat transfer coecient ratios of nanouid versus corresponding base uid
Eq. (10)

Experiment

hEF#1-1
hBF#1

50
70

1.19
1.19

1.03
1.03

hEF#1-2
hBF#1

50
70

1.36
1.36

1.22
1.15

hEF#2-1
hBF#2

50
70

1.02
1.02

1.01
1.01

hEF#2-2
hBF#2

50
70

1.14
1.14

1.08
1.07

14
12
10
8
6
4
2
0

14
12
10
8
6
4
2
0

Nu

Nu

Heating uid
temperature (C)

special factors in the nanouid mixture are aecting the

heat transfer mechanisms, especially for Series 1 uids at
higher wall temperature. The results suggest that heat
transfer property improvements can be made by varying
the base uid, the nanoparticle source, and the nanoparticle loading.
Nanoparticles may enhance the heat transfer coecient by two mechanisms: an increase in the thermal
conductivity of the overall system, and the movement
of the nanoparticles relative to the streamlines. Higher
temperature dierentials between the process uid and
the heat source result in lower heat transfer coecients
than expected. As the process uid begins to be heated
at the entrance of the exchanger, the uid viscosity decreases, the velocity prole ceases to be parabolic and

EF#1-1, 50C
0

50

100

150

200

EF1-1, 70C
0

50

100

Gz
12

14
12
10
8
6
4
2
0

10
6
4
2

EF#1-2, 50C

0
0

50

100

150

EF#1-2, 70C
0

200

50

Gz
14
12
10
8
6
4
2
0

EF#2-1, 50C
0

50

100

150

200

250

16
14
12
10
8
6
4
2
0

Nu
50

100

Gz

200

50

100

150

200

250

Gz

EF#2-2, 50C
0

150

EF#2-1, 70C

Gz
14
12
10
8
6
4
2
0

100

Gz

Nu

Nu

200

Nu

Nu

Nu

150

Gz

150

200

14
12
10
8
6
4
2
0

EF#2-2, 70C
0

50

100

150

200

Gz

Fig. 6. Relationship between the Nusselt and Graetz numbers for the test uids. () test data; (- - - -) calculation based on Oliver
correlation [19]; ( ) calculation based on Eubank and Proctor correlation [20].

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

starts to change. As shear rates change (the most change

will occur near the wall), nanoparticles may align with
the velocity. Alignment of the nanoparticles can decrease the thermal conductivities of uids if the nanoparticle-nanoparticle interactions responsible for the
enhanced heat transfer are disrupted. The shear force
in the uid or near the tube wall will make the particles
orient to some extent, and break up the percolation
structure. Higher shear rates can disperse nanoparticles
that have clustered, especially in a turbulent ow.
However, in a laminar ow system, this dispersing eect
might be relatively small. These experiments have a laminar ow (5 < Re < 110) with relatively high shear rate
(between 100 s1 and 1000 s1), which may contribute
to the lower heat transfer coecients. Near wall particle
depletion is another possible reason for the phenomenon. Moreover, adding particles into a uid increases
the viscosity of the nal mixture, which may suppress
the natural convection of the system. In summary, due
to the possible mechanisms for nanoparticle movements
in laminar ows under heating, Eq. (2) may not be directly applicable to the nanouid system.
The correlation developed by Li and Xuan [16,17],
Eq. (5), is based on a laminar ow of the nanouids.
However, this correlation predicts heat transfer coecients that are much higher than the experimental values. The correlation developed by Li does not include
the aspect ratio of the particles. Spherical copper nanoparticles were used in their work, while, in our system,
the nanoparticles are disc-like graphite nanoparticles.
There is at least one order of magnitude dierence in
the aspect ratio of the particles between these two cases.
The correlations by Oliver [19] and Eubank and
Proctor [20] address natural convention that will occur
in a laminar ow. The relationship between the Nusselt
and Graetz numbers of the test uids is plotted in Fig. 6,
together with theoretical curves. It can be seen that the
data from the Series 1 uids t the Oliver model fairly
well, while the data from the Series 2 uids lie in between
the plots of these two models. This indicates that an improved correlation for heat transfer coecient in nanouid systems should be developed, which is the future
direction of this research project.

4. Conclusion
The convection heat transfer performance of the
graphite nanouids were studied in laminar ow
through a circular tube. The experimental results show
that the nanoparticles increase the heat transfer coecient of the uid system in laminar ow, but the increase
is much less than that predicted by current correlation
based on static thermal conductivity measurements.
The type of nanoparticles, particle loading, base uid
chemistry, and process temperature are all important

1115

factors to be considered while developing nanouids

for high heat transfer coecients. Further investigation
is needed to develop an appropriate heat transfer correlation for non-spherical nanoparticle dispersions.
Acknowledgments
The authors appreciate very much the continuous
support and encouragement by Dr. Fran Lockwood of
the Valvoline Company. ZGZ, WBA and GF would like
to thank Dr. Steve Choi and Dr. Wenhua Yu of Argonne National Laboratory for the instructive discussion
and tremendous help on the construction of the transient hot-wire rig and the heat transfer loop. The authors
acknowledge the comments and suggestions of reviewers, who help improve the content of this paper.
References
[1] S.U.S. Choi, Enhancing thermal conductivity of uid with
nanoparticles, in: D.A. Siginer, H.P. Wang (Eds.), Development and Application of Non-Newronian Flows, The
American Society of Mechanical Engineers, New York,
NY, FED-vol. 231/MD-vol. 66, 1995, pp. 99105.
[2] S. Lee, S.U.S. Choi, Application of metallic nanoparticle
suspensions in advanced cooling systems, in: Y. Kwon, D.
Davis, H. Chung (Eds.), Recent advances in solids/structures and application of metallic materials, The American
Society of Mechanical Engineers, New York, NY, PVPvol. 342/MD-vol. 72, 1996, pp. 227234.
[3] J.A. Eastman, U.S. Choi, S. Li, L.J. Thompson, S. Lee,
Enhanced thermal conductivity through the development
of nanouids, Materials Research Society Symposium
Proceedings, vol. 457, Pittsburgh, PA, 1997, pp. 311.
[4] S. Lee, S.U.S. Choi, S. Li, J.A. Eastman, Measuring
thermal conductivity of uids containing oxide nanoparticles, J. Heat Transfer 121 (1999) 280289.
[5] X. Wang, X. Xu, S.U.S. Choi, Thermal conductivity of
nanoparticle-uid mixture, J. Thermophys. Heat Transfer
13 (4) (1999) 474480.
[6] Y. Xuan, Q. Li, Heat transfer enhancement of nanouids,
Int. J. Heat Transfer Fluid Flow 21 (2000) 5864.
[7] Y. Xuan, W. Roetzel, Conceptions for heat transfer
correlations of nanouids, Int. J. Heat Mass Transfer 43
(2000) 37013707.
[8] J.A. Eastman, S.U.S. Choi, S. Li, W. Yu, L.J. Thompson,
Anomalously increased eective thermal conductivities of
ethylene glycol-based nanouids containing copper nanoparticles, Appl. Phys. Lett. 78 (6) (2001) 718720.
[9] H. Xie, J. Wang, T. Xi, Y. Liu, Thermal conductivity of
suspensions containing nanosized SiC particles, Int. J.
Thermophys. 23 (2) (2002) 571580.
[10] S.U.S. Choi, Z.G. Zhang, W. Yu, F.E. Lockwood, E.A.
Grulke, Anomalous thermal conductivity enhancement in
nanotube suspensions, Appl. Phys. Lett. 79 (14) (2001)
22522254.
[11] H. Xie, J. Wang, T. Xi, Y. Liu, F. Ai, Q. Wu, Thermal
conductivity enhancement of suspensions containing nanosized alumina particles, J. Appl. Phys. 91 (7) (2002) 45684572.

1116

Y. Yang et al. / International Journal of Heat and Mass Transfer 48 (2005) 11071116

[12] H. Xie, J. Wang, T. Xi, Y. Liu, F. Ai, Thermal conductivity of suspensions containing SiC particles, J. Mater. Sci.
Lett. 21 (2002) 193195.
[13] J. Bicerano, J.F. Douglas, D.A. Brune, Model for the
viscosity of particle dispersions, J. Macromol. Sci., Part
CPolym. Rev. 39 (4) (1999) 561642.
[14] R.L. Hamilton, O.K. Crosser, Thermal conductivity of
heterogeneous two-component system, IEC Fundam. 1
(1962) 182191.
[15] P. Keblinski, S.R. Philpot, S.U.S. Choi, J.A. Eastman,
Mechanisms of heat ow in suspensions of nano-sized
particles (nanouids), Int. J. Heat Mass Transfer 45 (2002)
855863.
[16] Q. Li, Y. Xuan, Experimental investigation on convective
heat transfer of nanouids, Gongcheng Rewuli Xuebao (J.
Eng. Thermophys.) 23 (6) (2002) 721723.
[17] Q. Li, Y. Xuan, Convective heat transfer and ow
characteristics of Cu-water nanouid, Sci. China, Series
E: Technol. Sci. 45 (4) (2002) 408416.

[18] E.N. Sieder, G.E. Tate, Heat transfer and pressure drop of
liquids in tubes, Ind. Eng. Chem. 28 (12) (1936) 14291435.
[19] D.R. Oliver, Eect of natural convection on viscous-ow
heat transfer in horizontal tubes, Chem. Eng. Sci. 17 (1962)
335350.
[20] C.C. Eubank, W.S. Proctor, Eect of natural convection
on heat transfer with laminar ow in tubes, M.Sc. Thesis in
Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, 1951.
[21] Y. Nagasaka, A. Nagashima, Absolute measurement of
the thermal conductivity of electrically conducting liquids
by the transient hot-wire method, J. Phys. E: Sci. Instrum.
14 (1981) 14351440.
[22] C.C. Monrad, J.F. Pelton, Heat transfer by convection in
annular spaces, Trans. Am. Inst. Chem. Eng. 38 (1942)
593611.
[23] P.J.A. Hartman Kok, S.G. Kazarian, C.J. Lawrence, B.J.
Briscoe, Near-wall particle depletion in a owing colloidal
suspension, J. Rheol. 46 (2) (2002) 481493.